Data ExpoNumaga is exhibiting at Data Expo, 9 & 10 September. Come visit our stand.Get your free ticket
Numaga.Built by Replikate
Solutions

What you gain control over

Cost control

Usage insightsSee who uses AI, for what and how much: per user, group, agent and application.User limitsA usage ceiling per user and group, not a blank cheque.TiersThe right access level per group, not the most expensive profile for everyone.Model choiceNot one model: the right model, for every prompt.

Compliance

EU AI ActEvery prompt checked against its risk class, before the model call.GuardrailsYour AI policy, enforced at the gate, configurable per organisation.Audit trailsEvery interaction logged, and retrievable when you need it.Prompt loggingThe full content of every interaction (question and answer) recorded.RBACYour team sees only what it is entitled to, end-to-end.SSOPeople sign in with the account they already have; access follows your directory.

The platform

ChatOne familiar chat window, answers from your own knowledge, with sources.AppFor your people: just a familiar, secure assistant.Knowledge baseConnect your data sources; the knowledge base syncs automatically.EmbeddingsSearch your own documents by meaning, through the same gateway, with cost reported separately.Coding agentsRun your own coding agent (Claude Code, opencode) against the Numaga gateway.Applications & APIBuild Numaga governance into your own applications, via the API.
Solutions
Control planeCompliancePricingBlogAbout us
Book a demo

15 July 2026

New in Numaga: usage insights

See how AI is taking hold across your organisation: who uses it, with which models and whether it stays within budget. Drill down to every user, group, agent and application, inside the same governance.

As of today, Numaga gives you a full picture of how AI is taking hold across your organisation: who uses it, what for, with which models and whether it stays within budget. No separate BI project and no export to a spreadsheet: it lives in the dashboard, inside the same governance as the rest of your AI traffic.

Why this was needed

AI spreads through an organisation faster than the visibility into it. Within a few weeks dozens of people are chatting, agents are running and integrations are attached. Then the question arrives: who actually uses this, what for, and what does it cost? Without an answer, you steer on gut feeling.

Numaga routes all AI traffic through a single gateway. So usage doesn’t have to be reconstructed from scattered invoices or provider portals: it is counted at the source, per person, per model, per day. Usage insights make that count visible, without any data leaving the environment (single-tenant, hosted in the Netherlands).

What you see

The overview opens with the numbers that matter first: how many tokens and requests are running, how many people actively use AI, what share of your organisation is on board (adoption) and how often a guardrail stepped in. Below that you see usage over time and the model mix: which model does the work, and which barely gets used.

You pick the window yourself: the last 7, 30 or 90 days. The trend shows whether adoption is growing or stalling; the model mix shows whether expensive models are being used where a lighter one would do.

Drill down to the person

Every number is clickable. Usage insights offer four angles: per user, per group (the teams from your own identity provider), per agent (a personal key for an automated workflow) and per application (an integration running on a shared team key). Each entry point has its own detail page with the same layout: usage over time, model mix, guardrail interventions and the most recent requests.

So you answer a concrete question (“why did usage spike last week?”) in a few clicks, down to the individual request.

Tokens, not a fog of euros

Usage is counted in tokens, not in estimated euros. That is deliberate: tokens are the unit you actually manage by, and they map onto the budget of each user tier. Every user gets a status (comfortably within budget, near the limit, or over it) so you adjust before a bill surprises you.

Embeddings we count separately, in their own overview: they run as thousands of small calls and are priced in real rates, not token budgets. Putting them next to ordinary chat conversations would be comparing apples and oranges; we wrote about that earlier.

Who this matters to

Usage insights are for anyone accountable for AI use: the administrator watching budgets, the data protection officer who wants to know where guardrails intervene, the executive tracking adoption. It is read-only and visible to administrators and read-only auditors: view without being able to change anything.

Usage insights are on for every Numaga environment. Curious what it shows about your organisation? Book a demo and we’ll show it live.

← All articles

Numaga.

A managed, sovereign AI platform.

A product ofReplikate
PlatformControl planeModel routingArchitecturePricing
ComplianceEU AI ActAudit & observabilityAssurance packManaged service
ReplikateAbout usAbout ReplikateContactPrivacyResponsible disclosure
Numaga is a product of Replikate · ISO 27001 · Dutch sovereign infrastructure · © 2026